Architecture and System Implementation of Statistical Confidentiality Management for Regulatory Data in an Oracle Data Warehouse

The main objective of this bachelor's thesis is to design and implement a technical solution for the fully automated application of statistical confidentiality rules to aggregated regulatory data within the AnaCredit project. The thesis focuses on the efficient algorithmization of the dominance rule (identifying the two largest entities) and the frequency rule (minimum number of entities in an aggregate) directly at the Oracle database level. By utilizing advanced analytic functions, the solution reliably prevents both direct and indirect identification of reporting entities. The proposed ELT architecture is orchestrated by Oracle Data Integrator, ensuring detailed process auditability and guaranteeing continuous data availability for end users through asynchronous table switching. In the final phase, the high throughput of the system is demonstrated on a massive dataset with over 21 million records, where the targeted deployment of optimization techniques, such as parallel execution, interval partitioning, and bitmap indexes, reduced the total calculation time by 80%. The proposed solution was expertly validated by the AnaCredit business owners and the data warehouse development team at the Czech National Bank and subsequently deployed into the production environment, where it ensures the calculation of statistical confidentiality for three analytical reports used by reporting commercial banks.